Empirical Bayes Assessment of Differential Validity.
Abstract
An important problem in the assessment of test validity arises when the population under consideration consists of several distinct subgroups distributed in varying proportions across a number of training sites or programs. Separate prediction equations for each group at each site may be required, particularly if there is evidence of differential validity. The authors describe and illustrate a technique, based on the empirical Bayes paradigm, which provides estimates of the different prediction equations even when the members of a particular group are sparsely scattered across sites so that unique least squares estimates are not available. There is no limit to the number of groups, sites, or explanatory variables that can be considered.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1981
- Accession Number
- ADP001295
Entities
People
- Douglas H. Jones
- Henry I. Braun
Organizations
- Educational Testing Service